Why "Samsung exits" does not simply mean "less capacity"
Three distinctions matter more than the exit itself.
- What is being exited is 2D SLC NAND, not NAND. Mature-node 2D SLC has low output per unit area and weaker margin than 3D TLC/QLC, so the majors prioritise HBM and 3D NAND. This is structural abandonment of a low-margin category, not an overall production cut.
- There are few takers, and expansion is slow. Volume automotive- and industrial-grade SLC NAND now comes from a short list — Macronix, Winbond, Puya, Dosilicon. Even announced expansion needs dedicated process lines; new build or conversion runs at least 6–12 months. Short-run supply elasticity is very low.
- Demand is not substitutable. SLC's 1 bit/cell architecture gives high reliability, wide temperature (−40 °C to 105 °C) and long endurance (>100k cycles). In automotive ECUs, ADAS event recorders, industrial PLCs and AI edge boxes, TLC/QLC cannot stand in. If it goes short, the programme stops.
The driver is therefore rigid supply contraction meeting rigid demand — not the exit announcement.
Why automotive NOR Flash leads this cycle
- Supply: 55 nm/65 nm mature capacity is being redirected toward HBM-adjacent logic and 3D NAND controllers. Net new NOR capacity is close to zero, and industry inventory has run at roughly 3–5 weeks against an 8–12 week comfort line.
- Demand: smart cockpit, ADAS and body control consume high-reliability, fast-read NOR at a rate 3–5× that of a conventional vehicle — with automotive qualification cycles of 6–18 months making that demand highly inelastic.
- Special grades are worst. Wide-temperature (−40 °C to 125 °C), higher-voltage, small-density (16 Mb–512 Mb) automotive NOR is now supplied almost entirely by Macronix and Winbond. China-based vendors have entered, but incomplete automotive qualification, unstable volume yield and limited delivery capability mean they do not yet form an effective substitute.
The data point worth keeping: the market moved before the news
Many buyers assume a domestic alternative equals safety. Tracking mainstream automotive NOR parts (for example MX25L25645G and W25Q256JV) shows otherwise:
| Window | Price index | Inventory |
|---|---|---|
| Early March → early May | +17.08% | −48.14% |
| After the 29 April news | +0.30% | −1.36% |
The move had already happened. Suppliers tightened quotes and slowed shipment well before late April — they were not reacting to a headline. Meanwhile the domestic alternatives have entered the same cycle: AEC-Q100 qualified but limited in volume scale, low channel inventory, cheap sources disappearing, and fewer quoting sources compressing negotiating room.
A domestic alternative can mitigate discontinuity risk. It does not guarantee price stability. If the alternative is itself in a tight cycle, cost and delivery risk rise with it.
Three signals that separate real shortage from news
- Has inventory on your specific part fallen for three consecutive weeks?
- Has the number of quoting sources dropped from ten-plus to three to five?
- Has the gap between lowest and average price narrowed to within 5%?
Only when all three appear together has real scarcity reached the spot market, rather than living in expectation.
Bottom line
This cycle is the product of rigid supply contraction, non-substitutable high-reliability demand, and alternatives that have not yet built an effective buffer. For a buyer that means three things: don't switch to an alternative blindly — verify its own inventory and price trend first; don't wait for across-the-board shortage, because the early price-up, inventory-down phase is the warning window; and don't be misled by "someone is filling the gap" — filling a gap is not the same as shipping volume.
